p on Glassy-Polymeric Shear Deformation

5 hours ago - ... Materials - New in 2018 · ACS Biomaterials Science & Engineering ... ACS Energy Letters · ACS Infectious Diseases · ACS Macro Letter...
0 downloads 0 Views 4MB Size
Article Cite This: Macromolecules XXXX, XXX, XXX−XXX

Effect of the Ratio lK/p on Glassy-Polymeric Shear Deformation Mechanisms Hong T. Nguyen and Robert S. Hoy* Department of Physics, University of South Florida, Tampa, Florida 33620, United States ABSTRACT: Using molecular dynamics simulations, we study how chain stiffness affects how glassy polymers deform under applied shear. Loosely entangled systems composed of flexible chains exhibit strong shear banding and subsequent strain softening whereas tightly entangled systems composed of semiflexible chains exhibit neither of these. For all systems, inflection points in stress−strain curves correspond to the onset of chain scission, but nonlinear strain hardening continues up to much larger strains. Tightly entangled systems build up considerable elastic energy before fracturing via chain scission. This causes their plastic flow to be far more heterogeneous, and they ultimately fail along significantly sharper fracture planes than their loosely entangled counterparts, which fail via chain pullout. We quantify these differences using modern plasticity metrics and relate them to chain-stiffnessdependent differences in segmental packing efficiency and interchain interpenetration. It appears that the additional stress transmission mechanisms provided by the greater covalent bond tensions present in tightly entangled systems act to delocalize strain and promote more homogeneous deformation than is found in loosely entangled systems, but only until chain scission begins.

1. INTRODUCTION Experimental studies comparing the drastically different mechanical responses of chemically different polymer glasses have a long history.1−5 Many of the most basic features, such as tighter entanglement promoting ductility while greater aging promotes brittleness, are reasonably well understood. Far less well understood, however, is how these trends relate back to systems’ molecular architecture and microscopic interactions as well as how they relate to differences in the character of systems’ deformation micromechanisms. One obstacle to obtaining better understanding of these issues is that although polymeric systems’ entanglement density ρe, melt plateau modulus G0N, and glassy strain hardening modulus GR all scale as (lK/p)3, where lK and p are respectively their Kuhn and packing lengths,6−11 and it is well-known that higher entanglement density stabilizes systems against strain localization and brittle failure,12−14 it is difficult to relate these trends directly to chain stiffness because variable chain thickness (i.e., the presence of bulky side chains in some polymers but not others) obscures the relationship between chain stiffness and ρe.15−18 Thus, although increasing chain stiffness usually increases both ρe and toughness,4,5,16 counterexamples exist; e.g., polystyrene is less entangled and more brittle than polycarbonate even though polystrene is the stiffer polymer. Another source of difficulty is that traditional theoretical approaches1,13,19,20 treat systems as spatially homogeneous, ignoring the structural heterogeneity that has since been shown to be a key feature controlling the mechanical response of both nonpolymeric21,22 and polymeric glasses.23−25 Thus predicting even basic material properties such as whether a given system © XXXX American Chemical Society

will be brittle or ductile remains a focus of intense current research interest,25−27 and theoretical design of materials with optimized mechanical responsein particular, materials which preferentially fail via homogeneous shearremains a largely unrealized goal.11,28 Recent years have seen a profusion of simulation studies of the micromechanisms of glassy plasticity. Multiple metrics characterizing the spatiotemporal structure of plastic events have been developed and successfully related (at a predictive level) to systems’ microstructure.21,22,29−34 However, while it has long been known that polymer glasses’ plastic-deformation micromechanisms depend in detail on their microstructure (both intrachain and interchain1,3), application of these modern plasticity metrics to improving our understanding of their nonlinear shear response has only just begun.35 In this paper, using molecular dynamics simulations of a standard coarse-grained bead−spring polymer model,36 we study how the shear response of model polymer glasses depends on chain stiffness for f ixed chain thickness. By varying lK/p with all other system properties held fixedan operation which is very difficult to achieve in experimentswe isolate the role played by this critical parameter. Our results accord with established views of how chain stiffness influences shear deformation by controlling both segment-scale packing efficiency and the tightness of the entanglement mesh,8,12,13 but go significantly beyond previous work by clarifying how Received: March 27, 2018 Revised: May 9, 2018

A

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

that are our primary interest, all systems are cooled from T = Tg + 0.1ϵ/kB to T = Tg − 0.1ϵ/kB at a very slow rate (|Ṫ | = 10−6/τ) before continuing cooling to T = 0 at |Ṫ | = 10−5/τ. This protocol produces fairly well-aged46 glasses that can exhibit significant strain softening47,48 and hence should readily shear band if the constituent chains’ intrinsic properties (i.e., their microscopic interactions) support this. In stage (iii), systems are sheared in the yz-plane at a constant engineering strain rate γ̇ = 2.5 × 10−4/τ. This rate is low enough to be in the quasistatic regime49 where stresses vary logarithmically with γ̇; lower strain rates do not qualitatively change the results presented below. To mimic experiments, the stress along the x-direction (σxx) is maintained at zero thoughout the deformation runs. This protocol allows volume changes during deformation, which is critical since dilatation has long been known to be critical in controlling shear banding in amorphous materials.50 All systems are strained up to and slightly beyond the fracture strain γfrac (the strain at which the shear stress σyz is maximized). In all stages, angular interactions between three consecutive beads along chain backbones are modeled using the standard potential39

these relate to systems’ heterogeneous microstructure and heterogeneous microscale plastic deformation.

2. MODEL AND METHODS We employ the widely used Kremer−Grest bead−spring model,36 which has been shown to capture many features of Table 1. Glass Transition Temperatures and Statistical Properties of Chains at T = 0a kbend/ϵ

kBTg/ϵ

C∞

Ne

0 2 4

0.37 0.47 0.60

1.74 5.31 6.89

70 15 14

a

The entanglement length Ne is obtained by primitive path analysis (PPA).10 These values of Ne are slightly smaller than those reported in the literature for kBT = ϵ,10,44 but are consistent with the idea that entanglements in the glassy state are inherited from the T ≃ Tg supercooled melt1 and reflect the fact that for our chosen Ub(θ) chains straighten with decreasing T (which in turn decreases Ne43).

glassy-polymeric mechanical response.25,37 Systems are composed of Nch = 500 linear chains of N = 600 monomers. Periodic boundary conditions are applied along all three directions of (initially) cubic simulation cells. All monomers have mass m and interact via the truncated and shifted LennardJones potential 12 ⎡ ⎛ a ⎞6 ⎤ ⎛ a ⎞12 ⎛ a ⎞6 ⎛ a ⎞ ⎢ ⎜ ⎟ ⎜ ⎟ ULJ(r ) = 4ϵ − −⎜ ⎟ +⎜ ⎟⎥ ⎝r ⎠ ⎢⎣⎝ r ⎠ ⎝ rc ⎠ ⎝ rc ⎠ ⎥⎦

Ub(θ ) = k bend(1 − cos(θ ))

where θi is the angle between consecutive bond vectors bi⃗ and b⃗i+1; here b⃗i = ri⃗ +1 − ri⃗ and ri⃗ is the position of bead i. Note that θ is zero and Ub is minimized for straight trimers. We consider three chain stiffnesses ranging from flexible (kbend = 0) to fairly stiff (kbend = 4ϵ). Most previous studies have focused on kbend ≤ 2ϵ.25,37,42,51 Our kbend = 4ϵ chains are well into the semiflexible regime where lK is comparable to the tube diameter dT, producing entanglement that is qualitatively different from that found in flexible chains.52 The chain stiffnesses considered here are too low to produce long-range nematic order. However, short-range nematic order can also strongly affect mechanical properties. We characterize such short-range order via the methods employed in ref 53. The order parameter S is given by

(1)

where ϵ is the intermonomer binding energy, a is monomer diameter, and rc = 27/6a is the cutoff radius. The Lennard-Jones time unit is τ = ma 2 /ϵ , and the MD time step employed is δt = τ/200. All molecular dynamics simulations are performed using LAMMPS.38 Simulations are conducted in three stages: (i) melt equilibration at T > Tg; (ii) slow cooling to T = 0; (iii) applied shear deformation at T = 0. In stage (i), covalent bonds connecting consecutive monomers along chain backbones are modeled using the FENE potential UFENE(r) = −(kR02/2) ln[1 − (r/R0)2], with the standard36 parameters k = 30ϵ/a2 and R0 = 1.5a. Systems are thoroughly equilibrated39 at constant (zero) pressure at T = Tg + 0.1ϵ/kB, with pressure controlled using a Nosé−Hoover barostat. Note that the lower-than-usual equilibration temperature employed here is chosen because doing so is necessary to properly equilibrate the large-scale chain structure (and the entanglement length Ne) for semiflexible chains.40,43 Values of Tg for all systems are given in Table 1. In stage (ii), the covalent bond potential is switched to a quartic form suitable for studies of fracture: Uqu(r ) = U0 + kq(r − R b)3 (r − R b − B2 )

(3)

S=

3 Tr(q ̅ : q ̅ ) 2

where qαβ =

bα̂ bβ̂ −

1 δαβ 3 (4)

Here Tr is the trace operator, the ⟨...⟩ denote averaging over all cubic subcells of side length ∼3a and averaging of all normalized bond vectors b̂ = b⃗/|b⃗| in each such subcell, and bα̂ and bβ̂ are Cartesian components of b̂. In eq 4, S = 0 (S = 1) correspond to covalent bonds being perfectly unaligned (perfectly aligned) within each subcell. We find that in the glassy state S ranges from ≃0.3 for flexible chains to ≃0.45 for kbend = 4ϵ chains, with only a very weak T-dependence. This indicates thatas expectedneighboring covalent backbones are more aligned for stiffer chains. However, the magnitude of this effect is small. It only becomes large when midrange nematic order (order over distances ≳ 2a) develops, i.e., for kbend ≳ 5ϵ.53 Thus, we claim that the differences in systems’ mechanical response reported below have little to do with differences in their nematic order. We have also verified that all systems remain amorphous; i.e., increasing kbend does not produce any long-range positional order. We quantitatively characterize plastic deformation during stage (iii) using Falk and Langer’s nonaffinity measure D2min.29 For each monomer i, D2min is defined as the local deviation from

(2)

Bonds break when their length exceeds Rb = 1.3a. Following ref 45, B2 is selected by matching Uqu(r) to the first zero and the minimum of UFENE, which sets B2 = −0.4668a. Motivated by experimental data and previous bead−spring studies of glassypolymeric fracture,37,41 the ratio of the forces at which covalent and van der Waals bonds break is set to 100 by setting kq = 9640ϵ/a4. In order to minimize the effect of slow dynamics near the glass transition on the low-T mechanical properties B

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules affine deformation of the monomer’s “neighborhood” (the set of all ni monomers j such that |rj⃗ − ri⃗ | < rc) over the strain interval [γ ̅ − Δγ,̅ γ], ̅ i.e., over the tensorial strain increment Δγ:̅ n

−1 2 i Dmin, i = min[(ni ∑ j = 1 ∑α [ rji⃗ (γ ̅ ) · uα̂

−∑β (δαβ + Δϵαβ )( rji⃗ (γ ̅ − Δγ ̅ ) ·uβ̂ )])2]Δϵ̅

(5)

Here α and β denote the Cartesian directions x, y, z, ûα is the unit vector along direction α, and rj⃗ i = rj⃗ − ri⃗ . Δγ ̅ is the strain occurring over the applied shear strain increment Δγ = 5 × 10−3; this increment is small enough that multiple plastic rearrangements of a single region rarely occur. As in ref 29, plastic deformation is quantified by finding the tensorial strain increment Δϵ̅ that minimizes D2min,i. In general, Δϵ̅ ≠ Δγ,̅ and the degree of nonaffinity has been shown in many studies (e.g., refs 22, 29, 32, and 35) to correlate well with the degree of plastic activity.

3. RESULTS Figure 1a shows the stress−strain curves σyz(γ) for all systems. Results at small strains are consistent with many previous studies; an elastic response at small strains is followed by yielding, plastic flow, and strain hardening that becomes more dramatic (nonlinear) as γ increases.40 This degree of nonlinearity is far greater for the more tightly entangled semiflexible-chain systems due to the same factors discussed elsewhere (e.g., in refs 54−56); stretching of chains between entanglements both stores energy and increases plastic activity. The σyz(γ) curves have inflection points (switch from concaveup to concave-down) at characteristic strains γinfl(kbend) (Table 2) that decrease strongly with increasing chain stiffness.42 At larger strains, stress continues to increasealbeit with negative d2σyz/dγ2until systems fracture at γ ≃ γfrac. As expected, values of γfrac also decrease strongly with increasing chain stiffness. Note that the slow stress decrease shown for γ > γfrac is an artifact of the periodic boundary conditions (PBCs) employed here. PBCs are well-known to artificially suppress catastrophic failure, but eliminating this effect by making systems nonperiodic along one or more directions produces spuriously large surface effects for all currently computationally feasible system sizes. Thus, we focus on γ ≲ γfrac for the remainder of this paper. As expected,42,51 the inflection points in σyz(γ) correspond closely to the onset of bond scission. Figure 1b shows the fraction of broken covalent bonds f brok(γ) for the same systems. f brok(γ) is small for γ < γinfl and then rises sharply until systems fracture. Figure 1b also shows that the ultimate fraction of broken bonds grows sharply as chain stiffness increases and that the qualitative behavior of f brok(γ) in kbend ≥ 2ϵ systems is different from its behavior in flexible-chain systems. For kbend ≥ 2ϵ, the bond scission rate df brok/dγ peaks at γ ≃ γfrac, indicating that chain scission is the primary failure mode in these systems. In flexible-chain systems, no such peak in df brok/dγ is present. Instead, df brok/dγ increases at γ ≃ γfrac. We will show below that this occurs because these systems are failing via chain pullout. Figure 1c shows systems’ fractional volume change V(γ)/V0 − 1 during the deformation. All systems increase their volume during deformation by expanding along the transverse (x) direction. However, these increases differ strongly in both kind and degree. Dilatation has long been associated with shear banding.13,50 Here, flexible-chain systems dilate relatively rapidly at low strains. This stronger dilatation leads to the more dramatic yielding at γ = γyield shown in the inset of panel

Figure 1. Basic features of mechanical response: (a) shear stress σyz, (b) fraction of broken bonds, (c) fractional volume change, and (d) principal stress difference [σ1 − σ2] as a function of nominal shear strain γ for three different chain stiffnesses. Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. Vertical dotted lines indicate the onset of bond scission at γ = γinfl for each kbend, while vertical solid lines indicate the fracture strains γfrac.40 The insets to panels a and c highlight the response in the small-strain regime, and the dashed magenta curves in panel d show fits to the eight-chain model (eq 6). C

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

for γ ≲ γyield). Below we will relate these differences to differences in systems’ strain localization and fracture micromechanisms in much more detail. The eight-chain model of Arruda and Boyce57 has long been used to model glassy-polymeric deformation because it is mathematically simple yet can accurately capture stress−strain relations (including the nonlinear-hardening regime) for a wide variety of deformation modes.57,58 It predicts the stress difference between the principal stress components σi and σj to be

Table 2. Characteristic Strains γyield, γinfl, γfrac, and γcrit for Simple Shear Deformation at γ̇ = 2.5 × 10−4/τ40 kbend/ϵ

γyielda

γinfl

γfrac

γcrit (eq 8)

0 2 4

0.104 0.154 0.154

7.22 2.22 1.49

9.57 3.18 2.24

10.85 2.34 1.76

Values of γyield are identified by finding the strains at which dσyz/dγ is zero (for flexible chains) or locally minimal (for semiflexible chains).

a

Table 3. Fit Parameters to the Eight-Chain Model (Eq 6) Obtained from the Stress−Strain Curves Shown in Figure 1d kbend/ϵ

3 τflow 12 (ε/a )

GR (ε/a3)

Ne

0 2 4

1.745 1.274 1.261

0.223 1.197 1.176

136 22 16

σi − σj = τijflow + G R

3−1(h) 2 (λ i − λ j 2 ) 3h

(6)

where τflow is a plastic flow term, GR is the strain hardening ij modulus, λi and λj are the principal stretches along directions i and j, h = λchain/(Ne/C∞)1/2 where λchain = [(λ12 + λ22 + λ32)/ 3]1/2 describes the increase in chains’ mean end−end distance ⟨Ree2⟩1/2 during deformation, and 3−1 is the inverse Langevin function.57 Neglecting the small volume changes shown in panel c, the principal stretches are λ1 = ( γ 2 + 4 + γ )/2 , λ 2 = ( γ 2 + 4 − γ )/2 , and λ3 = 1. Stress is purely tensile along the λ1-axis and purely compressive along the λ2-axis; see Figure 3a for a schematic diagram of how these relate to our simulation geometry. In our simple shear deformation protocol, the stress tensor σ̅ has only four nonvanishing components: σyy, σzz, and σyz ≡ σzy. The two principal stresses (i.e., the stresses along the above-mentioned principal stretch directions 1, 2) are given by σ1,2 = (σyy + σzz)/2 ±

(σyy − σzz)2 /4 + σyz 2

(7)

Figure 1d shows the stress differences [σ1 − σ2](γ) together with fits to the eight-chain model for all systems. The fit parameters τflow 12 , GR, and Ne are given in Table 3. As in previous studies,9,11,55 the fit values of Ne are in general inconsistent with those obtained from rheological measurements and topological analyses. Nevertheless, if the fits are restricted to the strain range γyield < γ ≤ γinfl, eq 6 quite accurately captures the shape of the stress−strain curves. The dashed magenta curves in Figure 1d are fit over this range but are extended to larger γ in the plot to illustrate how the breakdown of the eight-chain model nicely coincides with the onset of bond scission. Since the character of the stresses in the different systems considered here become quite different for γ ≫ γyieldfor example, the stresses associated with bond tension grow large in tightly entangled semiflexible-chain systems but not in loosely entangled flexible-chain systems56the success of the eightchain model57,58 in capturing the shapes of stress−strain curves over such a wide range of γ is rather remarkable. The three systems considered here fracture at very different stress and strain levels. One way of comparing them on a more equal footing is to plot their mechanical response in appropriately scaled fashion. Figure 2a shows the shear stresses σyz scaled by their maximum values σmax = σyz(γfrac) and plotted against the scaled shear strain γ/γcrit. Here γcrit is the strain at which initially Gaussian chains of characteristic ratio C∞ and entanglement length Ne (with uniformly spaced entanglements and no chain slippage during deformation) pull taut. Its value for each system is given by the solution to the equation

Figure 2. Scaled stress−strain curves. Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. (a) The same results as in Figure 1a, but with strains scaled by γcrit(kbend) (eq 8; Table 2) and stresses scaled by the maximal values σmax yz = σyz(γfrac). (b) The corresponding results for crazed systems where crazing is induced by uniaxial-strain extension. For this protocol, γ is uniaxial stretch, and γcrit = 10.9, 2.5, and 2.0 for kbend = 0, 2ϵ, and 4ϵ.

a; we will show below that these trends are associated with formation of a “sharp” shear band. In contrast, intermediateflexibility (kbend = 2ϵ) systems deform at nearly constant volume prior to yielding, but dilate more strongly in the strainhardening regime, while stiffer (kbend = 4ϵ) chains exhibit much weaker dilatation. These differences are consistent with the lack of clear yielding and strain softening in these systems and suggest that their deformation is more homogeneous (at least

λ12(γcrit) + λ 2 2(γcrit) + λ32(γcrit) = 3Ne/C∞ D

(8)

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

Figure 3. Spatial structure of plastic activity: (a) The shear geometry and principal strain directions. Note that the angle between the direction (λ̂1) and the shear (xy) plane is α = tan−1(2/γ), while λ̂1 and λ̂2 remain perpendicular for all γ.5 (b−e) The 5% most damaged (highest D2min) monomers for (b) kbend = 0, γ = γyield; (c) kbend = 4ϵ, γ = γyield; (d) kbend = 4ϵ, γ = γfrac; (e) kbend = 0, γ = γfrac. Panel e is rotated to illustrate the fact that in there is no “fracture plane” of highly damaged atoms spanning the system. In panels b−e, monomers belonging to different chains are assigned different colors. Images were created using VMD.61

for the above-mentioned principal stretches (λ1, λ2, λ3). The scaled stresses for tightly entangled, semiflexible-chain systems evolve similarly with increasing scaled strain; both systems fracture at γ ≃ 1.1γcrit. A uniformly cross-linked network of Gaussian chains would fracture at γ ≃ γcrit, shortly after chains pulled taut. Real networks can be deformed to slightly higher strains without fracturing because of strand dispersity,59,60 and in uncrosslinked glasses, further deformation is allowed by chain slippage. Note that the γcrit values predicted by eq 8 may underestimate the strain at which chain tension starts increasing ahead of fracture in kbend ≥ 2ϵ systems because the end−end vectors of successive entangled segments are finitely correlated in these tightly entangled systems, providing an additional stress-relief mechanism.52,60 This is a subtle issue because eq 8’s assumption that chains are Gaussian at chemical distances n = Ne is only valid when Ne ≫ C∞ (and as indicated in Table 1, this is not the case for our kbend ≥ 2ϵ systems). Nevertheless, it is clear that our kbend ≥ 2ϵ systems behave roughly like cross-linked networks; chains pull taut between entanglement points and then undergo scission. Flexible-chain systems, however, behave quite differently, fracturing at γ ≃ 0.8γcrit. This is consistent with the argument given abovethat flexible-chain systems fail via chain pullout rather than scissionand we will support it further below. First, however, we will compare the results for shear shown in Figure 2a to their counterparts for crazing. The competition between crazing and shear is critical to understanding glassypolymeric fracture and is closely associated with these systems brittle−ductile transition.1,4,28,37 Figure 2b shows scaled stress− strain curves for crazing simulations wherein systems are extended uniaxially along the y direction at a rate γ̇ = 2.5 × 10−4/τ (where now γ = Ly/L0y is the uniaxial stretch rather than the shear strain) with their transverse dimensions held constant.41 The larger values of γfrac/γcrit for crazing are consistent with experimental trends;4 they arise from crazing’s much more dilatative character, which allows for much greater chain slippage. More importantly in our present context, since the qualitative trends for craze failure are the same as they are for shearless-entangled systems fail via chain pullout, at considerably lower values of γ/γcritthey should also remain the same under more general deformation protocols where crazing and shear can compete.4,5,37

The chain-stiffness-dependent differences discussed above can be elucidated by visualizing how deformation does (or does not) localize during shear. Figure 3 shows snapshots of the 5% most damaged (highest D2min) monomers for flexible and kbend = 4ϵ systems at various strains. Panels b and c show results for γ = γyield. The flexible-chain system has a sharply defined shear band extending across the system, nearly parallel to the xy-plane. This shear band is clearly associated with the distinct yielding and strain softening shown in Figure 1a. In contrast, the tightly entangled kbend = 4ϵ system shows no such shear banding. Instead, its plastic damage is diffuse and appears to be concentrated in the low-modulus, low-activation-energy “soft spots” discussed in many recent studies.21,30,31,62−64 Panels d and e show results for γ = γfrac. The tightly entangled system fails at a sharply defined fracture plane. This highly damaged region possesses a much higher fraction of broken bonds than the surrounding, less damaged region. Excess bond scission in this region shortly prior to fracture causes more load to be borne by the remaining bonds, further increasing their scission rate and producing a local mechanical instability that leads to rapid catastrophic failure at the fracture plane. This picture is not novelit is consistent with the standard view of chain-scission-dominated fracture1but it provides a useful contrast to the very different behavior of loosely entangled systems. The much larger values of γcrit in loosely entangled systems allow chains to become far more aligned prior to the onset of bond scission. Greater chain alignment is well known to reduce interchain friction in entangled polymer melts (see e.g. refs 65 and 66), and recent work has suggested that it also increases the effective tube diameter.67 These factors in combination with these systems’ greater dilatation (Figure 1c) strongly favor chain pullout over chain scission. As illustrated in panel e, failure of flexible-chain systems is not localized along a fracture plane but instead is rather homogeneous. This is consistent with the fact that chain pullout is a much less violent process because it occurs through breaking of van der Waals rather than covalent bonds. As such, chain pullout under applied shear deformation does not appear to induce the sort of sharp mechanical instability that favors strong strain localization. A deeper understanding of such effects can be gained by examining how the statistical properties of plastic damage evolve during deformation (Figure 4). Panel a shows the mean E

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

Figure 5. Connections between heterogeneous structure and heterogeneous plasticity. Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. (a) Voronoi volume distributions P(v). (b) Nonaffinity distributions P[log(D2min)]. Both panels show results for γ = γyield (dotted curves), γ = γinfl (dashed curves), and γ = γfrac (solid curves); panel a also shows results for undeformed systems (dash-dotted curves).

strong strain-rate dependencies observed in experimental studies of shear fracture28 imply that these couplings are strong. Accounting for the fact that these systems’ plastic deformation is heterogeneous provides additional insight. Panel b shows the standard deviation ΔD2min(γ) = [⟨(D2min(γ))2⟩ − 2 2 ⟨Dmin (γ)⟩2]1/2. At low strains, ΔDmin (γ) roughly tracks ⟨D2min(γ)⟩ and σdiss (γ). At the onset of bond scission, however, yz ΔD2min(γ) begins increasing more sharply, indicating that plastic deformation is becoming more heterogeneous. This observation is strengthened by plotting the heterogeneity measure H(γ) = ΔD2min(γ)/⟨D2min(γ)⟩ (panel c). At low strains, H(γ) is much greater in loosely entangled systems, consistent with the strong shear banding shown in Figure 3b. However, H(γ) increases sharply upon the onset of chain scission at γ = γinfl. Violent recoil from chain scission naturally leads to morelocalized plastic damage and hence the higher H(γ) for γ > γinfl. This increase is far stronger for tightly entangled systems and continues through fracture. All trends are consistent with the more localized fracture observed in these systems (Figure 3d). Next, in the spirit of many recent simulation studies of both nonpolymeric and polymeric glasses,21,22,30−33,64 we attempt to relate the heterogeneity of these systems’ plastic shear response to their structural heterogeneity. Since the long-standing assumption50 that increasing local free volume eases shear deformation (and particularly shear banding) suggests that monomers with larger D2min should preferentially have higher

Figure 4. Statistical measures of plasticity. Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. (a) The mean nonaffinity ⟨D2min(γ)⟩ measured over strain intervals δγ = 0.005. (b) The standard deviation ΔD2min(γ) of this measure. (c) The heterogeneity H(γ) = ΔD2min(γ)/⟨D2min(γ)⟩ of this measure. As in Figure 1, vertical dotted lines indicate γ = γinfl while vertical solid lines indicate γ = γfrac. Note that the large “noise” in all panels is the result of the avalanche-like plasticity typical22,34 of glasses deformed at low T.

damage ⟨D2min(γ)⟩ for all systems. In all cases, ⟨D2min(γ)⟩ is roughly proportional to the macroscopic shear stress σyz(γ) and more closely proportional to its dissipative component σdiss yz (γ). The large fluctuations are present because the low-temperature limit we are considering here produces the avalanche-like plasticity common to all low-T glasses.22,34 Similar trends have been observed for other deformation protocols such as uniaxialstress and plane-strain compression.55 Note that these fluctuations strengthen as strain rate decreases but are strongly suppressed as temperature increases.68 It will be interesting in future work to see how such temperature- and rate-dependent effects couple to chain-stiffness effectsparticularly since the F

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

where f(v) → 0 for v ≪ ⟨v⟩ and for v ≫ v*. The specific functional form of f(v) is challenging to determine69 but is not critical here. For γ ≲ γyield, the distributions are close to Gaussian (small f(v)) and relatively narrow (small σv), albeit wider for semiflexible systems as expected from these systems’ greater packing frustration.70 At γ ≃ γyield, the distributions begin to broaden. In particular, the large-v contribution grows; f(v) and v* both increase. Surprisingly, the magnitudes of these increases for γ ≪ γinfl are comparable for flexible-chain and semiflexible-chain systems even though the former shear-band while the latter do not. Once bond scission begins, however, the distributions broaden considerably more for semiflexible systems. Specifically, the low-v cutoff of f(v) appears to soften, and v* increases. This broadening reflects the violence of the bond scission process; postscission recoil naturally tends to create monomers with both low and high Voronoi volumes. Note that scission, in addition to being more prevalent in semiflexible systems (Figure 1b), is also more violent in these systems because of the greater strain energy released in individual scission events. Figure 5b shows the distributions of D2min values at the characteristic strains γyield, γinfl, and γfrac (Table 2). These distributions have approximately log-normal form: 2 )= P(Dmin

exp( −Δ2 /2)/ 2π Δ2 2 )⟩ ⟨log(Dmin

⎛ [log(D2 ) − ⟨log(D2 )⟩]2 ⎞ min min ⎟ × exp⎜ − 2Δ2 ⎝ ⎠

Log-normal distributions often arise in amorphous plasticity, e.g., in the distributions of plastic-zone activation energies and plastic-event sizes.34,71 They are very “fat-tailed” and indicate deformation physics dominated by rare, large, avalanche-like slip events.34 In fits to the distributions plotted in panel b, Δ/ ⟨log(D2min)⟩ increases dramatically with γ as systems’ plasticity becomes increasingly heterogeneous. The natural question to ask, then, is whether the trends in P(v) and P(D2min) with increasing shear strain are correlated and how any such correlations couple to other chain-stiffness effects. We calculated the Pearson’s correlation coefficient C(D2min, v) = cov(D2min,v)/(ΔD2min Δv), where cov(D2min, v) is the covariance of fluctuations of D2min and v while ΔD2min and Δv are their respective standard deviations, as a function of γ. C(γ) increases with strain but remains small [C(γ) ≤ 0.1] for all systems over the entire range 0 ≤ γ ≤ γfrac. Thus local plastic activity correlates only weakly with local free volume. While this lack of correlation accords with a previous study64 that found cavitation under tensile deformation correlates only weakly with Voronoi volume, it seems inconsistent with traditional views19,54 of the role free volume plays in the shear deformation of glassy polymers. We also attempted to relate the differences in our systems’ mechanical responses to differences in other metrics such as spatial correlations of D2min

Figure 6. Additional measures of plasticity: Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. (a) Spatial correlations of D2min fluctuations (F(r): eq 11). (b) Topological correlations of D2min fluctuations (G(n): eq 12). (c) Topological correlations of covalent bond tension fluctuations (K(n): eq 13). In all panels, dotted, dashed, and solid curves respectively show results for γyield, γinfl, and γfrac.

Voronoi volumes v before they rearrange, a particularly obvious spatially heterogeneous measure to examine is the Voronoi volume distribution. Figure 5a shows the distributions of Voronoi volumes P(v) in these systems. As is typical for glasses,69 the distributions are roughly fit by ⎛ (v − ⟨v⟩)2 ⎞ ⎛ v ⎞ ⎟ + f (v) exp⎜ − ⎟ P(v) = A exp⎜ − 2 ⎝ v* ⎠ 2σv ⎠ ⎝

(10)

F (r ) =

2 2 2 2 ⟨Dmin ( ri ⃗)Dmin ( ri ⃗ + r ⃗)⟩ − ⟨Dmin ⟩ 2 2 2 2 ⟨(Dmin ) ⟩ − ⟨Dmin ⟩

correlations of fluctuations in G(n) =

(9) G

D2min

(11)

along chain backbones

2 2 2 2 ⟨Dmin ( ri ⃗)Dmin ( ri +⃗ n)⟩ − ⟨Dmin ⟩ 2 2 2 2 ⟨(Dmin ) ⟩ − ⟨Dmin ⟩

(12) DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules and correlations of covalent-bond-tension fluctuations along chain backbones K (n) =

⟨T (bi⃗ )T (bi⃗ + n)⟩ − ⟨T ⟩2 ⟨T 2⟩ − ⟨T ⟩2

(13)

(where T = −∂[ULJ(r) + Uqu(r)]/∂r) that have proven useful in characterizing plastic deformation.22,37,56 In eqs 11−13, n is the chemical distance and brackets indicate averaging over all monomers i. Results are shown in Figure 6 and are consistent with expectations from previous studies. For example, results for F(r) show approximately exponential decay [F(r) ≃ F0 exp(−r/ ζ)] in the plastic-flow regime, where the length scale ζ corresponds to the typical avalanche size and F0 is maximal at γ ≃ γyield as is typical of amorphous materials.22,34 Results for G(n) and K(n) show increasing correlation on scales up to n ∼ Ne as systems approach fracture.41,56 These correlations have a clear albeit gradual crossover between slow decay for n ≲ Ne and faster decay for n ≳ Ne; note that the crossover is gradual rather than sharp because entanglement spacings are broadly distributed in undeformed melts and glasses60 and then evolve during deformation.72,73 Both spatial and topological measures of plasticity [i.e., F(r) and G(n)] are clearly much more correlated in tightly entangled systems. This is not surprising; it merely indicates the much greater importance of long-range intrachain stress transmission mechanisms (e.g., covalent bond tension) in these systems. However, it is useful to contrast the γ-dependencies of these measures. Tightly entangled systems’ D2min is maximally spatially correlated at γ ≃ γyield but maximally topologically correlated at γ ≃ γfrac. This distinction indirectly supports the notion that polymer glasses yield when their “primary network”26,27,74 of short-ranged van der Waals interactions fails but that in ductile polymers the entanglement network stabilizes further relatively homogeneous deformation until ultimate failure via chain pullout and scission.1,2

4. DISCUSSION AND CONCLUSIONS In this paper, we examined how chain stiffness affects shear deformation mechanisms in model polymer glasses using modern plasticity metrics. We isolated the effects of chain stiffness from competing chain thickness effects14,16 by studying a model in which thickness is fixed and the critical parameter lK/p monotonically increases with increasing stiffness. Our key findings were as follows: (i) flexible-chain systems are more prone to localized yielding via shear banding, while semiflexiblechain systems exhibit more spatially homogeneous yielding behavior; (ii) flexible-chain systems fail gradually and homogeneously via chain pullout, while semiflexible-chain systems fail violently at well-defined fracture planes via chain scission; (iii) the heterogeneity of systems’ plastic deformation increases sharply at the onset of bond scission; (iv) although free-volume- and nonaffine-deformation-based measures of plasticity both evolve dramatically during deformation, they are not strongly correlated to each other; (v) long-range stresstransmission in tightly entangled systems makes their plasticity more spatially and topologically correlated and thus suppresses strain localization. Findings (i) and (ii) generally accord with the established-consensus view of how chain stiffness influences shear deformation by setting the tightness of the entanglement mesh;8,12,13 here we have formulated (ii) more precisely by showing how the onset of chain scission triggers the mechanical

Figure 7. Effects of chain length, preparation protocol, and finite temperature on the stress−strain curves σyz(γ). Red, blue, and green respectively indicate data for systems with kbend/ϵ = 0, 2, and 4. Solid curves show the same results reported in Figure 1a. The dashed curves in panel a show results for flexible N = 40Ne systems. Dotted curves show results for (a) shorter, N = 8.6Ne chains; (b) rapidly quenched systems cooled to T = 0 at |Ṫ | = 10−3/τ; and (c) systems deformed at T = 3Tg/4 (where Tg values are given in Table 1). In all panels, insets highlight the small-strain response.

instability that produces strain localization and ultimately leads to fracture. The most novel aspect of our work is that we have related these more-macroscopic trends directly to findings iii− v, which are more microscopic (i.e., have smaller characteristic length scales) and are thus more difficult to observe in experiments. Taken together, our results indicate that the additional stress transmission mechanisms provided by the greater covalent bond tensions present in tightly entangled systems act to delocalize strain and promote more homogeH

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules

semiflexible counterparts. Note that the large N/Ne values at which the long-chain limit is reached in this study probably owe much to the Kremer−Grest model’s low monomeric friction coefficient.36,40 Conclusively determining how the phenomena discussed herein depend on chain length will require studying a wide range of model (or better yet, real) polymers and is beyond our scope. Another factor which is well-known13,46 to strongly affect polymer glasses’ mechanical response is their preparation protocol. Systems that were prepared by rapid thermal quenches or have been mechanically rejuvenated generally display minimal strain softening and maximal ductility, while systems prepared by slow cooling from the melt followed by significant aging below Tg display opposite trends.14,25 The preparation protocol described in section 2 produces the bestaged systems readily attainable given current computer power. Panel b contrasts their response to that of systems obtained via rapid (|Ṫ | = 10−3/τ) quenches from T = ϵ/kB to T = 0. Differences in the mechanical response at small strains are typical of those observed in experiments;13,14 quenched systems strain-soften less and exhibit more diffuse shear bands than their annealed counterparts. At larger strains, differences in σyz(γ) are considerably smaller, consistent with the common view that yielding rejuvenates glasses. Interestingly, however, some differences persist further beyond yield. For example, ⟨D2min(γ)⟩ values remain larger for the fast-quenched systems; their smaller σ/⟨D2min⟩ ratios reflect the fact that less-aged glasses sit higher on their energy landscapes and are locally “softer”.30,31,46,76,77 These differences may have broad implications for the character of plastic deformation in these systems. For example, ref 78 concluded that the cooperativity of plastic deformation processes in polymer glasses scales with ρe based on the different strain-rate dependencies of the mechanical responses of PS−PPO blends with different ρ e . The preparation-protocol dependence of σyz(γ) illustrated here is markedly stronger for kbend = 0 systems than for their semiflexible counterparts, suggesting that it will be of particular interest to test the ideas of ref 78 by comparing the γ̇ dependencies of the mechanical responses of bead−spring systems (with different lK/p) prepared using the same protocol. Finally, polymer glasses’ mechanical response does of course vary strongly with temperature. While a detailed consideration of thermal effects is beyond the scope of this paper, we demonstrate here that results obtained for the T = 0 limit are not irrelevant to typical ambient conditions. Panel c contrasts the T = 0 results to those for T = 3Tg/4, a typical value of Troom/Tg for many synthetic polymers.25 At the high strain rate considered here (γ̇ = 2.5 × 10−4/τ), differences are quantitative rather than qualitative. For example, greater thermal activation facilitates chain pullout and reduces bond scission, which in turn increases γfrac, but the strength of these effects does not strongly couple to chain stiffness. Future work will examine thermal and rate effects in greater detail.

neous deformation than is found in loosely entangled systems, but only until chain scission begins. Many of the chain-stiffness-dependent differences reported herein may be understood in terms of corresponding differences in segmental packing. Larger values of lK/p correspond to greater interchain interpenetration on the segmental scale. Since our flexible-chain systems have larger elastic moduli owing to their more efficient segmental packing,3,70,75 yet less interchain interpenetration, they are naturally more prone to shear banding. In contrast, our semiflexible-chain systems’ less efficient packing and greater interchain interpenetration make them dilate less at small strains. Combined with their intrinsically stronger strain hardening response, this stabilizes them against shear banding. At larger strains, flexible-chain systems’ lesser interchain interpenetration facilitates their ultimate failure via chain pullout, while semiflexible-chain systems’ tighter entanglement (associated with their larger lK/p7,8,10) favors their more violent chain-scission-dominated failure. All these results are consistent with Haward’s idea12 that the dominant parameter controlling polymer glasses’ large-strain response is the density ρc ∼ (lK/ p)3 of uncrossable chain contours. Here we focused on slowly cooled, relatively well-aged46 glasses that occupy the lower regions of their energy landscapes prior to deformation. We focused on the low-temperature limit of these glasses’ mechanical response because plasticity is simplest to interpret when it is purely strain-activated. Future work will examine how thermal and thermal-history effects couple to the chain-stiffness effects discussed herein, as well as how the micromechanisms of shear deformation through fracture couple to microscale structural features such as elastic heterogeneity21,30,31,62−64 and/or “softness”.22,33 A brief discussion of these issues and some preliminary results are given in the Appendix.



APPENDIX. SENSITIVITY OF RESULTS TO CHAIN LENGTH, TEMPERATURE, AND PREPARATION PROTOCOL It has long been known that chain length is a critical parameter determining polymer glasses’ mechanical properties. Under typical experimental conditions, short chains (N ≲ 10Ne) favor failure via chain pullout, while long chains (N ≳10Ne) favor failure via chain scission.1 Above, we considered N = 600 systems which respectively have N/Ne = 8.6, 40, and 43 for kbend/ϵ = 0, 2, and 4. We implicitly assumed that we were considering the long-chain limit. To check the validity of this assumption, we tested the response of systems with N/Ne = 8.6, i.e., with N = 600, 128, and 120 for kbend/ϵ = 0, 2, and 4, and also of kbend = 0 systems with N/Ne = 40 (N = 2800). All simulations were conducted using the same protocols described in section 2. Stress−strain curves for all these systems are compared in Figure 7a. As expected, results for all chain lengths are near-identical at small and moderate strains. Only in the dramatic hardening regime where energetic contributions to stress become significant (and especially so for tightly entangled systems56) does significant N-dependence of σyz(γ) appear. Longer-chain systems show a greater degree of strain hardening and ultimately fracture more violently. This is expected since chain pullout is suppressed in these systems.42 Critically, however, differences between the responses of the N/Ne = 40 systems are qualitatively the same as those discussed in section 3; flexible systems still exhibit a more pulloutdominated, homogeneous fracture phenomenology than their



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (R.S.H.). ORCID

Robert S. Hoy: 0000-0003-0283-8117 Notes

The authors declare no competing financial interest. I

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules



Deformation of a Poly(methyl methacrylate) Glass. Macromolecules 2014, 47, 800−806. (25) Roth, C. B., Ed.; Polymer Glasses; CRC Press: 2016. (26) Wang, S.-Q.; Cheng, S. W.; Lin, P. P.; Li, X. X. A phenomenological molecular model for yielding and brittle-ductile transition of polymer glasses. J. Chem. Phys. 2014, 141, 094905. (27) Li, X. X.; Wang, S.-Q. Mapping brittle and ductile behaviors of polymeric glasses under large extension. ACS Macro Lett. 2015, 4, 1110. (28) Archer, J. S.; Lesser, A. J. Shear band formation and mode II fracture of polymeric glasses. J. Polym. Sci., Part B: Polym. Phys. 2011, 49, 103. (29) Falk, M. L.; Langer, J. S. Dynamics of viscoplastic deformation in amorphous solids. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 1998, 57, 7192. (30) Manning, M. L.; Liu, A. J. Vibrational modes identify soft spots in a sheared disordered packing. Phys. Rev. Lett. 2011, 107, 108302. (31) Smessaert, A.; Rottler, J. Correlation between rearrangements and soft modes in polymer glasses during deformation and recovery. Phys. Rev. E 2015, 92, 052308. (32) Nicolas, A.; Rottler, J.; Barrat, J. Spatiotemportal correlations between plastic events in the shear flow of athermal amorphous solids. Eur. Phys. J. E: Soft Matter Biol. Phys. 2014, 37, 50. (33) Cubuk, E. D.; Schoenholz, S. S.; Rieser, J. M.; Malone, B. D.; Rottler, J.; Durian, D. J.; Kaxiras, E.; Liu, A. J. Identifying structural flow defects in disordered solids using machine-learning methods. Phys. Rev. Lett. 2015, 114, 108001. (34) Uhl, J. T.; et al. Universal quake statistics: from compressed nanocrystals to earthquakes. Sci. Rep. 2015, 5, 16493. (35) Shavit, A.; Riggleman, R. A. Strain localization in glassy polymers under cylindrical confinement. Phys. Chem. Chem. Phys. 2014, 16, 10301. (36) Kremer, K.; Grest, G. S. Dynamics of entangled linear polymer melts - a molecular dynamics simulation. J. Chem. Phys. 1990, 92, 5057. (37) Rottler, J. Fracture in glassy polymers: a molecular modeling perspective. J. Phys.: Condens. Matter 2009, 21, 463101. (38) Plimpton, S. Fast parallel algorithms for short-range molecular dynamics. J. Comput. Phys. 1995, 117, 1. (39) Auhl, R.; Everaers, R.; Grest, G. S.; Kremer, K.; Plimpton, S. J. Equilibration of long chain polymer melts in computer simulations. J. Chem. Phys. 2003, 119, 12718. (40) The increases in strain hardening and fracture stress and the decrease in fracture strain with increasing kbend and increasing N/Ne shown in Figures 1, 2, and 7 are stronger than in previous studies37,41,42 because we equilibrate systems at T = Tg + 0.1ϵ/kB rather than equilibrating at T = 1.0ϵ/kB as was the common practice prior to 2016.43 (41) Rottler, J.; Robbins, M. O. Growth, microstructure, and failure of crazes in glassy polymers. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 2003, 68, 011801. (42) Ge, T.; Robbins, M. O.; Perahia, D.; Grest, G. S. Healing of polymer interfaces: Interfacial dynamics, entanglements, and strength. Phys. Rev. E 2014, 90, 012602. (43) Grest, G. S. Communication: Polymer entanglement dynamics: Role of attractive interactions. J. Chem. Phys. 2016, 145, 141101. (44) Hoy, R. S.; Foteinopoulou, K.; Kröger, M. Topological analysis of polymeric melts: Chain-length effects and fast-converging estimators for entanglement length. Phys. Rev. E 2009, 80, 031803. (45) Sliozberg, Y. R.; Hoy, R. S.; Mrozek, R. A.; Lenhart, J. L.; Andzelm, J. W. Role of entanglements and bond scission in high strainrate deformation of polymer gels. Polymer 2014, 55, 2543. (46) Strum, L. C. E. Physical aging in plastics and other glassy materials. Polym. Eng. Sci. 1977, 17, 165. (47) Warren, M.; Rottler, J. Simulations of aging and plastic deformation in polymer glasses. Phys. Rev. E 2007, 76, 031802. (48) Rottler, J.; Warren, M. Deformation, yield and ageing in glassy solids. Eur. Phys. J.: Spec. Top. 2008, 161, 55.

ACKNOWLEDGMENTS Mark O. Robbins, Ralf Everaers, and Gary S. Grest provided helpful discussions. This material is based upon work supported by the National Science Foundation under Grant DMR1555242.



REFERENCES

(1) Kramer, E. J. Microscopic and molecular fundamentals of crazing. Adv. Polym. Sci. 1983, 52-53, 1. (2) Vincent, P. I. The tough-brittle transition in thermoplastics. Polymer 1960, 1, 425−44. (3) Wellinghoff, S. T.; Baer, E. Microstructure and its relationship to deformation processes in amorphous polymer glasses. J. Appl. Polym. Sci. 1978, 22, 2025. (4) Henkee, C. S.; Kramer, E. J. Crazing and shear deformation in crosslinked polystyrene. J. Polym. Sci., Polym. Phys. Ed. 1984, 22, 721− 737. (5) G’Sell, C.; Boni, S.; Shrivastava, S. Application of the plane simple shear test for determination of the plastic behavior of solid polymers at large strains. J. Mater. Sci. 1983, 18, 903. (6) Fetters, L. J.; Lohse, D. J.; Richter, D.; Witten, T. A.; Zirkel, A. Macromolecules 1994, 27, 4639. (7) Fetters, L. J.; Lohse, D. J.; Graessley, W. W. Chain dimensions and entanglement spacings in dense macromolecular systems. J. Polym. Sci., Part B: Polym. Phys. 1999, 37, 1023. (8) Fetters, L. J.; Lohse, D. J.; Milner, S. T.; Graessley, W. W. Packing length influence in linear polymer melts on the entanglement, critical, and reptation molecular weights. Macromolecules 1999, 32, 6847. (9) van Melick, H. G. H.; Govaert, L. E.; Meijer, H. E. H. On the origin of strain hardening in glassy polymers. Polymer 2003, 44, 2493. (10) Everaers, R.; Sukumaran, S. K.; Grest, G. S.; Svaneborg, C.; Sivasubramanian, A.; Kremer, K. Rheology and microscopic topology of entangled polymeric liquids. Science 2004, 303, 823. (11) Kramer, E. J. Open questions in the physics of deformation of polymer glasses. J. Polym. Sci., Part B: Polym. Phys. 2005, 43, 3369. (12) Haward, R. N. Strain-hardening of thermoplastics. Macromolecules 1993, 26, 5860. (13) Haward, R. N., Young, R. J., Eds.; The Physics of Glassy Polymers, 2nd ed.; Chapman and Hall: London, 1997. (14) Meijer, H. E. H.; Govaert, L. E. Mechanical performance of polymer systems: The relation between structure and properties. Prog. Polym. Sci. 2005, 30, 915. (15) Wu, S. Chain structure and entanglement. J. Polym. Sci., Part B: Polym. Phys. 1989, 27, 723. (16) Wu, S. Chain structure, phase morphology, and toughness relationships in polymers and blends. Polym. Eng. Sci. 1990, 30, 753. (17) Wang, S. On chain statistics and entanglement of linear flexible polymer melts. Macromolecules 2007, 40, 8684. (18) Matsumiya, Y.; Watanabe, H. Nonlinear Stress Relaxation of Miscible Polyisoprene/Poly(p-tert- butylstyrene) Blends in Pseudomonodisperse State. Macromolecules 2016, 49, 4544. (19) Anand, L.; Gurtin, M. E. A theory of amorphous solids undergoing large deformations, with application to polymeric glasses. Int. J. Solids Struct. 2003, 40, 1465. (20) Klompen, E. T. J.; Engels, T. A. P.; Govaert, L. E.; Meijer, H. E. H. Modeling of the postyield response of glassy polymers: Influence of thermomechanical history. Macromolecules 2005, 38, 6997. (21) Schoenholz, S. S.; Liu, A. J.; Riggleman, R. A.; Rottler, J. Understanding plastic deformation in thermal glasses from single-softspot dynamics. Phys. Rev. X 2014, 4, 031014. (22) Cubuk, E. D.; et al. Structure-property relationships from universal signatures of plasticity in disordered solids. Science 2017, 358, 1033. (23) Lee, H. N.; Riggleman, R. A.; de Pablo, J. J.; Ediger, M. D. Deformation-Induced Mobility in Polymer Glasses during Multistep Creep Experiments and Simulations. Macromolecules 2009, 42, 4328. (24) Bending, B.; Christison, K.; Ricci, J.; Ediger, M. D. Measurement of Segmental Mobility during Constant Strain Rate J

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX

Article

Macromolecules (49) Rottler, J.; Robbins, M. O. Shear yielding of amorphous polymer glasses: effect of temperature and strain rate. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 2003, 68, 011507. (50) Gilman, J. J. Mechanical behavior of metallic glasses. J. Appl. Phys. 1975, 46, 1625. (51) Ge, T.; Pierce, F.; Perahia, D.; Grest, G. S.; Robbins, M. O. Molecular dynamics simulations of polymer welding: strength from interfacial entanglements. Phys. Rev. Lett. 2013, 110, 098301. (52) Uchida, N.; Grest, G. S.; Everaers, R. Viscoelasticity and primitive path analysis of entangled polymer liquids: From F-actin to polyethylene. J. Chem. Phys. 2008, 128, 044902. (53) Nguyen, H. T.; Smith, T. B.; Hoy, R. S.; Karayiannis, N. C. Effect of chain stiffness on the competition between crystallization and glass- formation in model unentangled polymers. J. Chem. Phys. 2015, 143, 144901. (54) Hasan, O. A.; Boyce, M. C. A constitutive model for the nonlinear viscoelastic viscoplastic behavior of glassy polymers. Polym. Eng. Sci. 1995, 35, 331. (55) Hoy, R. S.; Robbins, M. O. Strain hardening of polymer glasses: Entanglements, energetics, and plasticity. Phys. Rev. E 2008, 77, 031801. (56) Hoy, R. S. Why is understanding glassy polymer mechanics so difficult? J. Polym. Sci., Part B: Polym. Phys. 2011, 49, 979. (57) Arruda, E. M.; Boyce, M. C. Evolution of plastic anisotropy in amorphous polymers during finite straining. Int. J. Plast. 1993, 9, 697. (58) Boyce, M. C.; Arruda, E. M.; Jayachandran, R. The large-strain compression, tension, and simple shear of polycarbonate. Polym. Eng. Sci. 1994, 34, 716. (59) Svaneborg, C.; Grest, G. S.; Everaers, R. Disorder effects on the strain response of model polymer networks. Polymer 2005, 46, 4283. (60) Tzoumanekas, C.; Theodorou, D. N. Topological analysis of linear polymer melts: A statistical approach. Macromolecules 2006, 39, 4592. (61) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular dynamics. J. Mol. Graphics 1996, 14, 33. (62) Smessaert, A.; Rottler, J. Structural relaxation in glassy polymers predicted by soft modes: a quantitative analysis. Soft Matter 2014, 10, 8533. (63) Riggleman, R. A.; Douglas, J. F.; de Pablo, J. J. Antiplasticization and the elastic properties of glass-forming polymer liquids. Soft Matter 2010, 6, 292. (64) Makke, A.; Perez, M.; Rottler, J.; Lame, O.; Barrat, J. L. Predictors of Cavitation in Glassy Polymers under Tensile Strain: A Coarse-Grained Molecular Dynamics Investigation. Macromol. Theory Simul. 2011, 20, 826. (65) Wang, S.-Q.; Ravindranath, S.; Wang, Y.; Boukany, P. New theoretical considerations in polymer rheology: elastic breakdown of chain entanglement network. J. Chem. Phys. 2007, 127, 064903. (66) Yaoita, T.; Isaki, T.; Masubuchi, Y.; Watanabe, H.; Ianniruberto, G.; Marrucci, G. Primitive chain network simulation of elongational flows of entangled linear chains: Stretch/orientation-induced reduction of monomeric friction. Macromolecules 2012, 45, 2773. (67) Schweizer, K. S.; Sussman, D. M. A force-level theory of the rheology of entangled rod and chain polymer liquids. I. Tube deformation, microscopic yielding, and the nonlinear elastic limit. J. Chem. Phys. 2016, 145, 214903. (68) Salerno, K. M.; Maloney, C. E.; Robbins, M. O. Avalanches in Strained Amorphous Solids: Does Inertia Destroy Critical Behavior? Phys. Rev. Lett. 2012, 109, 105703. (69) Kumar, V. S.; Kumaran, V. Voronoi cell volume distribution and configurational entropy of hard-spheres. J. Chem. Phys. 2005, 123, 114501. (70) Kumar, R.; Goswami, M.; Sumpter, B. G.; Novikov, V. N.; Sokolov, A. P. Effects of backbone rigidity on the local structure and dynamics in polymer melts and glasses. Phys. Chem. Chem. Phys. 2013, 15, 4604. (71) Derlet, P. M.; Maass, R. Linking high- and low-temperature plasticity in bulk metallic glasses II: use of a log-normal barrier energy

distribution and a mean-field description of high-temperature plasticity. Philos. Mag. 2014, 94, 2776. (72) Leonforte, F. Evolution of entanglements during the response to a uniaxial deformation of lamellar triblock copolymers and polymer glasses. Phys. Rev. E 2010, 82, 041802. (73) Ge, T.; Tzoumanekas, C.; Anogiannakis, S. D.; Hoy, R. S.; Robbins, M. O. Entanglements in Glassy Polymer Crazing: CrossLinks or Tubes? Macromolecules 2017, 50, 459. (74) Li, X.; Liu, J.; Liu, Z.; Tsige, M.; Wang, S. Illustrating the Molecular Origin of Mechanical Stress in Ductile Deformation of Polymer Glasses. Phys. Rev. Lett. 2018, 120, 077801. (75) Ness, C.; Palyulin, V. V.; Milkus, R.; Elder, R.; Sirk, T.; Zaccone, A. Nonmonotonic dependence of polymer-glass mechanical response on chain bending stiffness. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 2017, 96, 030501. (76) Stillinger, F. H. A Topographic View of Supercooled Liquids and Glass-Formation. Science 1995, 267, 1935. (77) Rodney, D.; Schroder, T. On the potential energy landscape of supercooled liquids and glasses. Eur. Phys. J. E: Soft Matter Biol. Phys. 2011, 34, 100. (78) Ho, J.; Govaert, L. E.; Utz, M. Plastic deformation of glassy polymers: correlation of shear activation volume and entanglement density. Macromolecules 2003, 36, 7398.

K

DOI: 10.1021/acs.macromol.8b00651 Macromolecules XXXX, XXX, XXX−XXX